Using a Convolutional Neural Network to Classify Nanobodies on Neutralizing Ability Using Sequence Data for SARS-Cov-2


  • NITHYA VINODH School of Systems Biology, George Mason University, Fairfax, VA
  • losif Vaisman School of Systems Biology, George Mason University, Fairfax, VA



In order to combat the growing number of viral diseases, scientists have found a variety of new technologies to help reduce the risk of contracting a disease. Of the many that have been found over the last few years, one of the most promising are nanobodies. Nanobodies, much like antibodies, are able to bind to viruses. However, they consist only of a heavy chain variable domain, making them significantly smaller than antibodies and far more stable (Bao et al., 2021). They are also easier to produce synthetically, making them a viable alternative to current monoclonal antibody therapies. Certain nanobodies have the ability to neutralize antigens by binding to the outer regions of the virus and preventing it from entering cells and replicating. However, there are no ways to predict whether a given nanobody would neutralize a given virus without expensive wet lab testing, which is expensive and time consuming, Bioinformatics, however, provides tools that can speed up the discovery of neutralizing nanobodies. While methods exist to determine nanobody structure based on sequence, none exist to classify nanobodies based on neutralization ability. To address this, a CNN was created in order to classify nanobodies' neutralizing abilities using only sequence data for SARS-Cov-2. This yielded a preliminary model with a 69.4% accuracy in predicting the neutralization activity of nanobodies against SARS-Cov-2, which indicates with further testing, an accurate model to determine neutralization activity can be created.





College of Science: School of Systems Biology